Multilayer Clustering: Biomarker Driven Segmentation of Alzheimer’s Disease Patient Population

Identification of biomarkers for the Alzheimer’s disease is a challenge and a very difficult task both for medical research and data analysis. In this work we present results obtained by application of a novel clustering tool. The goal is to identify subpopulations of the Alzheimer’s disease (AD) patients that are homogeneous in respect of available clinical and biological descriptors. The result presents a segmentation of the Alzheimer’s disease patient population and it may be expected that within each subpopulation separately it will be easier to identify connections between clinical and biological descriptors. Through the evaluation of the obtained clusters with AD subpopulations it has been noticed that for two of them relevant biological measurements (whole brain volume and intracerebral volume) change in opposite directions. If this observation is actually true it would mean that the diagnosed severe dementia problems are results of different physiological processes. The observation may have substantial consequences for medical research and clinical trial design. The used clustering methodology may be interesting also for other medical and biological domains.